ADVENTURES IN SONIC RESTORATION WITH THE CEDAR CR I AND DC I
MIX MAGAZINE, October 1994, by George Petersen

A rather interesting project came along recently, and I just couldn't resist taking on a CD reissue series of early jazz and pop 78 rpm discs recorded from 1909 to 1949. The studio has a number of restoration tools available, such as notch filter sets, parametric and program (solidstate and tube types) equalizers, compressors, expanders, gates, a Burwen dynamic noise filter, a multichannel disk-based editing system, sample rate converters (hardware and software-based), various analog and digital tape-recording formats and a shiny
box of single-edged razor blades sitting next to the 2-track. However, a number of interesting new devices for audio restoration have recently come to market, and the 78 reissue project seemed the perfect opportunity to check these out.

Founded in 1988 in Cambridge, England, Cedar (Computer Enhanced Digital Audio Restoration) Audio Ltd. began by building PC-based systems for cleaning up sound stored on tape, vinyl and film. Cedar has developed a series of stand- alone 2-channel devices for audio restoration applications. Housed in a two-
rackspace chassis, the devices feature 16-/24-bit digital I/O, 40-bit internal processing and battery-backed RAM for storage of setups and user parameters. All operations happen entirely in the digital domain, although the rack units also include balanced and unbalanced analog I/O (with 16- and 18-bit, 64-times oversampling delta-sigma bit-stream converters) for applications where interfacing with existing analog systems is required. All of the units operate in near real time, with a slight processing delay in the lOOms to 200ms range.

The first product in this series is the Cedar DC 1 De-Clicker ($16,500), a digital system featuring a unique process to remove clicks and scratches without removing transients in the program material. Input signals pass
through the DC 1 unaffected, and the processing is only applied when a click or scratch occurs. Once a click is detected, that section is removed and replaced with a high-order interpolation of what should have been there, based on the waveform before and after the click. The microprocessor within the DC 1 is capable of 50 million floating-point calculations per second for removing up to 5,000 scratches per second.

The CR 1 De-Crackler ($19,500) uses the same split and recombine process as the Cedar-2 PC-based system, whereby the elements of a recording that contain crackle, buzz or distortion are separated from the unaffected segment of the input signal; after processing the degraded section, both segments of the signal are recombined into a crackle-free output.

DC I AND CR 1 OPERATIONS

Setup is equally straightforward just make the required analog (-10 dB unbalanced RCA or +4dB balanced SLR) or digital (AES/EBU or S/PDIF coaxial) connections, and you're on your way. The analog and digital outputs are always active, so coming in digital and leaving as analog (or vice versa) is no problem. Inputs and outputs can be monitored using the tenets LED ladder meters, but I found these to be of almost no help at all. The ballistics are too slow, and the bottom of the scale is only -23 dB, which doesn't provide much of an idea of what's happening to signals below that threshold. So before you can react to a hot input, it's already lit the red peak LED.

There's a massive amount of circuitry packed into these two-rack-space enclosures, requiring an internal (rear) cooling fan. In a quiet control room, the fan noise may be audible, but it's certainly no problem and is barely
noticeable, especially when the units are mounted in a rack.

At this stage, I could have routed the preamp directly through the DC 1 and CR 1 units and into the D-10 in the digital domain, but I wanted to have an unprocessed archive tape of the 78s, and each disc required slightly some more than slightly! different settings on the DC 1 and CR 1. So rather than playing each disc five or six times to find the optimum Cedar settings, I could transfer the archive tapes through the DC 1 and CR 1 in the digital domain at a later date, and tweak at my leisure.

Everything was recorded on two channels, using a stereo cartridge. Although the first experimental stereo records were made in 1932, none of the discs in this particular project was originally recorded in stereo. However, even on this mono project, Cedar's stereo processing capability came in handy. By
recording both sides of the record groove as discrete channels, I could later choose the better-sounding of the two. About 80% of the time, one channel is audibly better than the other. In some cases, this would change over the course of a single disc, and once the

THE PROJECT

I began my restoration project by manually cleaning the records and auditioning cuts to find the best versions in cases where several discs were available. I didn't use modern phono preamps, as these incorporate the RIAA
equalization curve, which doesn't match prewar 78s. The phono outputs from the Esoteric Sound Ramses turntable were routed into two channels of an outboard mic preamp.

Each disk was matched to its original playback speed, accomplished by cross-referencing the date of the recording and comparing the performance to concert pitch using a Kurzweil piano module in the control room. I wish somebody would make a reference 78 disc with calibration recordings were transferred into a
disc-based system, I could cut and paste between the two channels to create a single seamless mono track.

The next step was transferring the archived recordings through the Cedar boxes in the digital domain. By the way, the Cedar units support 32, 44.1 and 48kHz sampling rates from digital inputs, and 44.1 or 48 kHz can be selected when inputting analog sources.

Then the fun really began. Thanks to the user interface, tweaking the Cedar units is an intuitive and simple process, which emphasizes creativity and speed, as opposed to bogging down the operator with multilevel submenus.

The DC 1 De-Clicker has settings for small, medium or large clicks (scratches), with each representing a different processing algorithm. The small setting handles small ticks with a short duration and medium deals with longer clicks that require more interpolation. However, the longer setting not only provides increased interpolation for larger, longer duration click, but also compensates for the LF resonance such as those introduced by the tone arm that the click creates, occurring after the large click. The other main control of the DC 1 is its threshold control, adjustable in 20 steps to specify the size of the scratch removed by the process.

The CR. 1 De-Crackler is designed to remove crackle, buzz and distortion from many sources, whether created by record damage, overloaded analog inputs or distortion on ma tape or optical film soundtracks. Adjusting the CR. 1 is a bit more complicated than using the DC 1. The split level is determined by
detect level controls, and by manipulating these, the user can decide how much processing is required.

The CR. 1 provides two crackle algorithms to choose from: Crackle 1 is for well defined crackle, and Crackle 2 is intended for duller or grungy-sounding crackle. The operational threshold (defining when the processing kicks in) is based on a 0 to 40 scale, although typical threshold settings are in the 4 to 5 range. I would have preferred a scale where the typical settings fell somewhere in the middle of the range, thus providing a wider latitude of adjustment. On most of the material I was restoring, the threshold settings ranged from 2 to 6, and I sometimes wanted a median setting, such as 3.5, when 3 seemed too low and 4 may have been too high. Perhaps Cedar could remedy this in a future software update.

It didn't take long for me to appreciate the real-time aspect of the CR. 1 and DC 1. Questionable material (due to poor condition) could be auditioned immediately to determine its usability, rather than pouring it into a
disabused system, waiting for processing and then finding it to be unusable. This real-time advantage of the Cedar is an obvious plus in broadcasting situations, or in circumstances where materials are Cedar processed before being transferred into a computer editing system. Even if you're pouring a project into a system such as the Sonic Solutions nonuse workstation, which has substantial restoration capabilities, pre-processing through the DC 1 and CR. 1 can save time by handling much of the clean-up chores before the next stage begins. Of course, its possible to remove scratches on most disabused editors via cut-and-paste editing or waveform-redraw techniques, but there's no contest when comparing manual editing to Cedars ability to remove hundreds or thousands of scratches per second.

One thing to keep in mind about using the DC 1 and CR. 1 is that these boxes can handle about 80% of the restoration chores on vintage records. This left me with the relatively simple job of adding a comparatively gentle touch of mastering EQ, which consisted of a narrow band notch filter at 6.8 kHz (to eliminate some residual crackling) and some wide band program shaping with a tube equalizer, for overall tonal contouring.

OTHER APPLICATIONS

In addition to 78 rpm record restoration, the Cedar units are equally useful on cylinders and other formats including LPs, 45s and 16 rpm transcription records. Results on film soundtracks were mixed. The system was ineffective on older, variable density optical tracks as clicks and pops on such tracks aren't really transient enough for the Cedar to recognize without interfering with the program material. The DC 1 and CR. 1 worked much better on variable area optical (which have been in general use since the mid-1940s), where the shape of the click and pop wave forms tends to be sharper and more pointed, and thus easier to detect. In sum, the effect on soundtracks was less pronounced than on records but still a noticeable improvement. Overall, Cedars
DC 1 and CR. 1 provide an impressive amount of flexibility in an easy-to-use system. The attention to audio throughout is obvious from the whisper clean reproduction to the lack of artifacts or harshness in the output signal. To be sure, these are not inexpensive, but they are well crafted, powerful tools for the serious audio restorer or archivist.

Return to Magazine Reviews

d_prevpg.gif (6066 bytes)